author’s postprint version of the article on institutional
TRANSCRIPT
1
Author’s postprint version of the article on institutional repository 1
after 12 months embargo from first online publication 2
(Published online: 30 June 2015) 3
4
5
Spatial, seasonal trends and transboundary transport of PM2.5 6
inorganic ions in the Veneto region (Northeastern Italy) 7
8
Mauro Masiola
, Francesca Benetellob 9
Roy M. Harrisona†
, Gianni Formentonc 10
Francesco De Gasparic, Bruno Pavoni
b 11
12
Published in 13
Atmospheric Environment 117 (2015) 19-31 14
15
16
http://dx.doi.org/10.1016/j.atmosenv.2015.06.044 17
18
https://www.sciencedirect.com/science/article/abs/pii/S13522310119
5301862?via%3Dihub 20
21
22
23
24
25
Released under a Creative Commons Attribution Non-Commercial No 26
Derivatives License 27
28
29
30
31
32
To whom correspondence should be addressed. Email: [email protected]
† Also at: Department of Environmental Sciences / Center of Excellence in Environmental Studies, King
Abdulaziz University, PO Box 80203, Jeddah, 21589, Saudi Arabia
2
SPATIAL, SEASONAL TRENDS AND 33
TRANSBOUNDARY TRANSPORT OF PM2.5 34
INORGANIC IONS IN THE VENETO 35
REGION (NORTHEASTERN ITALY) 36
37
Mauro Masiola
, Francesca Benetellob 38
Roy M. Harrisona†
, Gianni Formentonc 39
Francesco De Gasparic, Bruno Pavoni
b 40
41
42 aDivision of Environmental Health and Risk Management 43
School of Geography, Earth and Environmental Sciences 44
University of Birmingham 45
Edgbaston, Birmingham B15 2TT 46
United Kingdom 47
48 bDipartimento di Scienze Ambientali 49
Informatica e Statistica, Università Ca’ Foscari Venezia 50
Dorsoduro 2137, 30123 Venezia, Italy 51
52 cDipartimento Provinciale di Padova 53
Agenzia Regionale per la Prevenzione e Protezione Ambientale 54
del Veneto (ARPAV), Via Ospedale 22, 35121 Padova, Italy 55
56 57 58
59
To whom correspondence should be addressed. Email: [email protected]
† Also at: Department of Environmental Sciences / Center of Excellence in Environmental Studies, King
Abdulaziz University, PO Box 80203, Jeddah, 21589, Saudi Arabia
3
ABSTRACT 60
The Veneto Region lies in the eastern part of the Po Valley (Italy). This is one of the hotspots 61
in Europe for air quality, where efforts to meet the European standard for PM2.5 according to 62
current and future legislation have been generally unsuccessful. Recent data indicating that 63
ammonium, nitrate and sulphate account for about one third of total PM2.5 mass show that 64
secondary inorganic aerosol (SIA) plays a key role in the exceedence of the standards. A 65
sampling campaign for PM2.5 was carried out simultaneously in six major cities (2012-2013). 66
The water soluble inorganic ions were quantified and data processed to: (1) investigate the 67
seasonal trends and the spatial variations of the ionic component of aerosol; (2) identify 68
chemical characteristics at the regional-scale and (3) assess the potential effects of long-range 69
transport using back-trajectory cluster analysis and concentration-weighted trajectory (CWT) 70
models. Results indicated that PM2.5 and SIA ions have an increasing gradient in 71
concentrations from North (mountain) to South (lowland) and from East (coastal) to West 72
(more continental), whereas K+ and Ca
2+ levels are quite uniformly distributed. Similar 73
seasonal trends in PM2.5 and ions are seen across the region. Simultaneous daily changes 74
were observed and interpreted as a consequence of similar emission sources, secondary 75
pollutant generation and accumulation/removal processes. Sulphate and nitrate were not 76
directly related to the concentrations of their precursor gases and were generally largely, but 77
not completely, neutralised by ammonium. The clustering of back-trajectories and CWT 78
demonstrate that the long-range movement of the air masses has a major impact upon PM2.5 79
and ion concentrations: an area spreading from Eastern to Central Europe was identified as a 80
main potential source for most ions. The valley sites are also heavily influenced by local 81
emissions in slow moving northerly air masses. Finally, two episodes of high nitrate levels 82
were investigated to explain why some sites are experiencing much higher concentrations 83
than others. This study identifies some key features in the generation of SIA in the Po Valley, 84
4
demonstrating that SIA generation is a regional pollution phenomenon and mitigation 85
policies are required at regional, national and even European scales. 86
87
Keywords: PM2.5, Ionic composition, Secondary inorganic aerosol, Long-range transport, Po 88
Valley 89
5
1. INTRODUCTION 90
Although most elements of the periodic table and many thousands of different organic 91
compounds are found in airborne particulate matter (PM), a few major components usually 92
make up a large percentage of the total mass. Ammonium (NH4+), nitrate (NO3
–) and sulphate 93
(SO42–
) are among the major components of aerosol in the lower troposphere and their 94
average mass percentages in fine PM (aerodynamic diameter less than 2.5 µm, PM2.5) 95
account for ~7%, ~9% and ~15%, respectively in southern Europe (Putaud et al., 2010). 96
These ions can be directly emitted from various sources, including sea salt, mineral dust, 97
traffic, biomass combustion, industries and other anthropogenic processes. However, the 98
dominant mechanisms for their presence in the particulate-phase are the oxidation of 99
precursor gases, i.e. nitrogen oxides (NO+NO2=NOx) and sulfur dioxide (SO2), to nitric 100
(HNO3) and sulfuric (H2SO4) acids, respectively. The subsequent neutralisation with 101
ammonia (NH3) forms salts such as ammonium nitrate (NH4NO3), ammonium sulphate 102
((NH4)2SO4) and ammonium bisulphate ((NH4)HSO4) (Seinfeld and Pandis, 2006; Holmes, 103
2007; Benson et al., 2011). These salts are commonly referred to as secondary inorganic 104
aerosol (SIA). 105
106
PM2.5 has clearly demonstrated adverse effects upon human health (WHO, 2006), and 107
reducing human exposure to PM is, therefore, of primary importance. In particular, it is a key 108
objective in the few hot-spots left in Europe, such as the Po Valley, where the current 109
standards for PM are not met. Several large cities (e.g., Milan, Turin, Bologna, Verona and 110
Venice-Mestre) and a myriad of minor urban agglomerations, industrial areas, agricultural 111
and rural environments are spread over a ~48·103 km
2-wide alluvial lowland. A total of ~16 112
million inhabitants and the related road traffic and energy production cause heavy 113
anthropogenic emissions across the entire valley. In addition, enclosure by the Alps and 114
6
Apennine mountains surrounding the valley from the North, West and South (only the eastern 115
side is opened to the Adriatic Sea) forms a barrier for the dispersion of pollutants and has a 116
negative impact on air quality, with a buildup of PM and nitrogen oxides mainly during the 117
cold season. Sampling at a rural site in the south-eastern Po Valley (San Pietro Capofiume), 118
Decesari et al. (2014) found that sulphate and nitrate contributed appreciably to particulate 119
matter mass. Their analysis of the association of particulate matter concentrations with 120
meteorological factors revealed a complex interplay of local and long-range transport 121
influences. 122
123
The European Directive 2008/50/EC imposed a PM2.5 annual average concentration of 25 µg 124
m–3
as a target value to be achieved by 2010. As the target value will become the European 125
limit value to be met by 2015, this standard has to be achieved with the current and future 126
legislation. However this concentration is not met in many locations of Veneto Region, 127
Eastern Po Valley (EEA, 2013): in 2012, eight of the 14 sites included in the main monitoring 128
plan for PM2.5 of the local environmental protection agency (ARPAV) breached the target 129
value (ARPAV, 2013). These sites are located in a number of major cities of the region and 130
generally the PM2.5 concentrations were 3−7 µg m−3
above the target value. In addition, PM2.5 131
levels exceeding the target value were also recorded in rural environments demonstrating that 132
even the background pollution is high. 133
134
Almost all the literature available for the SIA pollution in the Veneto is based on studies 135
carried out in the Municipality of Venice (Squizzato et al., 2012;2013; Masiol et al., 2014a). 136
Results have shown that about 25−35% of the total PM2.5 mass in Venice-Mestre is made up 137
of SIA, which is therefore a key component when the target values in the eastern Po Valley 138
are exceeded. Consequently, successful policies should include not only the reduction of 139
7
direct (primary) sources, but also the reduction of precursor gases to prevent the formation of 140
secondary particles (de Leeuw, 2002; Andreani-Aksoyoglu et al., 2004; Wu et al., 2008). 141
142
However, data collected in a single coastal city, Venice, are not sufficient to depict the key 143
characteristics of SIA pollution across the Veneto, the territory of which extends from Alpine 144
environments to foothills to flat plain areas in the North-South axis and extends from 145
continental to coastal environments in the West-East axis. 146
147
In view of this, the present study investigates the levels, spatial distribution and sources of 148
SIA in six major cities of Veneto, which have been carefully selected to be representative of 149
different environments of the region. The investigated territory extends to ~125 km on the 150
North-South axis and ~60 km from West to Est. The inorganic ionic composition of PM2.5 151
was quantified at six sites located in major cities for one year (2012−2013). The seasonal and 152
spatial variations were examined using a series of statistical tests and chemometric 153
approaches. Starting from the experimental data, the SIA formation at a regional-scale in 154
Veneto is described and the potential local and external sources are investigated. This study 155
has identified some key features that can improve the understanding of the generation of 156
secondary inorganic particles in the entire Po Valley. 157
158
2. MATERIALS AND METHODS 159
2.1 Site Selection 160
A multiple-site PM2.5 sampling campaign was carried out according to the EN 14907:2005 161
standard from April 2012 to March 2013 in 6 major cities: Belluno (BL), Conegliano (TV), 162
Vicenza (VI), Venice-Mestre (VE), Padua (PD) and Rovigo (RO) (Figure 1a). Stations 163
managed by ARPAV, were placed in high density residential areas and can be considered as 164
8
representative of city-wide background levels. In Table 1 some site characteristics are 165
summarised. Since the Veneto region includes a northern Alpine zone (29% of the territory), 166
an intermediate hilly one (15%), a wide southern flat lowland (56%) and an eastern coastline 167
(95 km long), the cities were also selected to represent most of the differing environments 168
and features of the territory. BL (36,600 inhabitants) is located in an Alpine valley 169
surrounded by mountains, with no large industries or heavy traffic, but biomass burning 170
emissions are intense in winter, as wood is largely used for domestic heating. TV (35,700 171
inhabitants) is in a foothill region and is therefore representative of the transition between the 172
mountain and lowland; many factories process stainless steel, produce appliances and 173
electrical equipment, but a large part of the land is used for agriculture, especially for 174
vineyards. VI (115,900 inhabitants) is an important city with intense traffic and small to 175
medium-sized mechanical, textile, tanning and jewelry manufactures. VE (271,000 176
inhabitants) is a conurbation extending from the coastal lagoon of Venice to the mainland 177
with a complex emission scenario. This includes heavy road, maritime and airport traffic, an 178
industrial zone hosting chemical and steel plants, an oil-refinery, incineration facilities, 179
thermoelectric power plants and others. PD (214,200 inhabitants) is the most densely 180
populated municipality of the region, with many medium-sized factories mainly in the 181
engineering, technological and building sectors, but it also suffers from intense traffic due to 182
the presence of a large intermodal and logistics hub. RO (52,800 inhabitants) is located in a 183
flat lowland midway between the Alps and the Apennines and is the biggest processing center 184
of Veneto for agricultural products. Demographic data refer to 2011 and to the whole 185
municipalities. 186
187
188
189
9
2.2 Experimental 190
PM was collected on quartz fiber filters (Whatman QMA), starting at midnight for 24 h 191
continuously using low-volume samplers installed in air conditioned cabins (temperature 192
<20°C). PM2.5 masses were gravimetrically determined (sensitivity 0.1 µg) after 193
preconditioning at constant temperature (20±1 °C) and relative humidity (50±5%). Sampled 194
filters were stored in clean Petri slides in the dark and at −20 °C until analyses to prevent 195
losses, photochemical reactions and biological processes. The entire set of collected samples 196
covers most of the year (total 2190). The quantification of the water soluble inorganic ions 197
was limited to a subset of 60 samples per site (total 360) collected in 6 periods of 10 198
consecutive days in the middle of April, June, August, October, December and February. 199
Periods were chosen to be representative of all the seasons and include the dates when home 200
heating was switched off (15 April) and on (15 October) as established by the national 201
legislation. A ~2 cm2-wide subsample of each filter was extracted in vials with 10 mL MilliQ 202
water (resistivity= 18.2 MΩ·cm at 25°C, Millipore) and sonicated for 50 min. Vials were 203
capped to avoid artifacts and sample evaporation. Extracts were pre-filtered on microporous 204
(0.45 µm) PTFE membranes and injected in two Metrohm (Switzerland) ion chromatographic 205
systems with conductivity detectors to quantify the concentrations of five anions (F−, Cl
−, 206
NO3−, PO4
3−, SO4
2−) and five cations (Na
+, NH4
+, K
+, Mg
2+, Ca
2+). Anions were separated on 207
a Metrosep A Supp 7–250/4.0 column applying a isocratic flow (0.8 mL min–1
) of 360 mM 208
Na2CO3 (Sigma-Aldrich, ACS ≥99.8%) eluent. Cations were determined using a Metrosep C 209
3–150/4.0 column and a 1 mL min–1
isocratic flow of 3 mM ultrapure HNO3 (Fluka, 210
TraceSELECT, ≥69%). Single-ionic standards were prepared from pure salts and used to test 211
the linearity and calibrate the instrumental responses. The analyses were routinely checked by 212
using certified liquid standards (Fluka, TraceCERT) diluted in MilliQ water. The relative 213
repeatability of each ion determination (standard deviation of 10 replications) was <5%. Field 214
10
blanks were prepared and analysed together with the samples and the values obtained were 215
routinely subtracted. Limits of detections (LODs) were calculated as three times the standard 216
deviation of field blanks: data below the LODs were substituted by LOD/2. 217
218
Other chemical parameters were automatically determined on hourly or bihourly basis in each 219
site following European standards: NO, NO2, NOx (EN 14211:2012); SO2 (EN 14212:2012); 220
O3 (EN 14625:2012); PM10 and PM2.5 with automatic beta-attenuation monitor systems. A 221
comprehensive list of measured parameters in each site is provided in Table 1. 222
223
2.3 Sampling Artifacts 224
A number of studies have reported that potential artifacts can occur during air sampling 225
because of ambient conditions and the interactions between collected particles and gaseous 226
compounds with each other or with the filter medium (e.g., Appel et al., 1984; Dasch et al., 227
1989; Harrison et al., 1990; Harrison and Kitto, 1990; Koutrakis et al., 1992; Zhang and 228
McMurry, 1992; Cheng and Tsai, 1997; Pathak et al., 2004a; Schaap et al., 2004a; Pathak and 229
Chan, 2005). Generally, the most evident artifact is the evaporation of nitrate due to its gas-230
particle partitioning (negative artifact), which is further enhanced by higher temperatures and 231
drier air. Also, pressure drop across the filter and mixing of acidic and alkaline particles on 232
the filter may perturb the gas-particle equilibrium. On the contrary, absorption of gas-phase 233
nitric acid may also occur (positive artifact) mainly driven by the presence of sea-salt 234
particles. 235
236
Studies conducted in the Po Valley (Putaud et al., 2002; Schaap et al., 2004a) have reported 237
that nitrate volatilization generally dominates over absorption. In particular, Schaap et al., 238
(2004a) concluded that quartz filters have a full retention of nitrate at temperatures <20°C. In 239
11
this study, all the samplers were installed into air conditioned cabins with an internal 240
constant temperature below 20°C. Subsequent filter transport, handling and analysis were 241
carried out under the same controlled conditions, while filter storage was at −20 °C. 242
Moreover, the prevailing high relative humidity recorded at all of the sites (average >70% 243
RH) during the sampling periods further decreased the potential nitrate loss. For these 244
reasons, negative artifacts of nitrate can be considered negligible. Positive artifacts are also 245
expected to be small: concentrations of Na+ and Cl
‒ (as tracers of sea-salt) and Mg
2+ and Ca
2+ 246
(as tracers of crustal particles) during the study were very low. 247
248
Another potential positive artefact can be caused by the absorption of SO2 on collected 249
particles, which can be subsequently oxidized to sulphate (Pathak and Chan, 2005). Due to 250
the very low concentrations of SO2 in Veneto (ARPAV, 2013) and according to data obtained 251
with and without the use of denuders by Vecchi et al. (2009), sulphate can be considered a 252
conserved specie in the Po Valley (i.e. not subject to adsorption or volatilisation). 253
254
In summary, sampling conditions and chemical results indicate that potential artifacts in this 255
study are small. For this reason, all the chemometric analyses have been performed on raw 256
data. 257
258
2.4 Back-Trajectory and CWT Analysis 259
Back-trajectories were computed to study the history of air masses during the sampling days. 260
Set-up: HYSPLIT model (Draxler and Rolph, 2013; Rolph, 2013); 96 h backward; starting 261
height at 20 m a.s.l.; 4 trajectories per day at 3, 9, 15 and 21 UTC calculated separately for all 262
the sites. A clustering algorithm using the Euclidean distance measure (Carslaw, 2014) was 263
12
applied to gain information on pollutant species with similar chemical histories by grouping 264
back-trajectories into clusters depending on their potential origin. 265
266
CWT is a back-trajectory-based hybrid receptor model used to assess potential source areas 267
affecting air pollution at a receptor site. Briefly, each grid cell ij in a grid domain was used to 268
compute the weighted concentration obtained by averaging sample concentrations that have 269
associated trajectories passing the grid cell according to: 270
𝐶𝑖𝑗 =1
∑ 𝜏𝑖𝑗𝑘𝑁𝑘=1
∑(𝐶𝑘)𝜏𝑖𝑗𝑘
𝑁
𝑘=1
where i and j are the coordinates of grid, k the trajectory index, N the number of trajectories, 271
Ck the pollutant concentration measured at the receptor site upon arrival of the trajectory k, 272
and τijk represents the residence time of trajectory k in the ij cell. Further insights are provided 273
in Seibert et al. (1994) and Hsu et al. (2003). Cluster analysis and CWT were computed using 274
R and the ‘Openair’ package (Carslaw and Ropkins, 2012; Carslaw, 2013). 275
276
3. RESULTS 277
3.1 Overview of Results 278
Table 2 summarises the annual average concentrations of PM2.5 and ions and also gives 279
statistics for SIA (as sum of ammonium, nitrate and sulphate) and ΣWSII (sum of all the 280
analysed water soluble inorganic ions). Due to the high percentage of samples below the 281
LODs, F−, Mg
2+ and PO4
3− were excluded from the statistics. A comprehensive list of results 282
for each month is provided as Supplementary Information Table SI1. The PM2.5 annual 283
average concentrations (365 days) ranged from a minimum of 16 µg m–3
in BL and a 284
maximum of 28 µg m–3
in PD. In the study period, the European annual average target value 285
of 25 µg m–3
(2008/50/EC Directive) was breached in three sites (PD, RO, VI). On an annual 286
13
basis, ΣWSII accounted for a significant fraction of the total PM2.5 mass, ranging from 30% 287
(BL) to 41% (RO) and generally showed a slightly increasing trend from north to south. 288
Annually, the most abundant ion in all the sites (Figure 1b) was nitrate, ranging from 36% 289
(BL) and 47% (VI) of the ΣWSII, followed by sulphate 22% (VI)−29% (BL, VE), 290
ammonium 17% (BL, TV)−21% (VI, RO) and potassium 3% (RO, VI)−5% (BL). Sodium 291
varied from 1% (VI) and 5% (TV), while the remaining single ions never exceeded 2%. The 292
annual levels of PM2.5 and PM2.5-bound nitrate, sulphate and ammonium in this study are 293
very similar to those recorded in other urban sites in the Po Valley (Table SI2). 294
295
Gaseous pollutants were recorded for all the year on hourly basis and data were averaged to 296
give daily mean values (Table 2 and Table SI1). The annual average concentrations of NO 297
during the selected periods varied from 12 µg m–3
(TV) to 27 µg m–3
(PD); NO2 from 23 µg 298
m–3
(BL) to 37 µg m–3
(PD and RO); NOx from 45 µg m–3
(BL and TV) to 79 µg m–3
(PD); 299
O3 from 46 µg m–3
(RO) to 61 µg m–3
(PD); SO2 from 1 µg m–3
(PD) to 2.8 µg m–3
(VE). 300
These mean concentrations are very close to the annual average levels and demonstrate that 301
the selected periods are representative of the annual concentrations. 302
303
The annual average NO2 levels never exceeded the Limit Value fixed by the European 304
Directives (40 µg m–3
). In Veneto the emission inventory (EI) for 2007/8 (ARPAV and 305
Regione Veneto, 2013) reported that road transport was the main source of NOx (52111 Mg 306
y-1
), followed by combustion in manufacturing industry (15119 Mg y-1
), other mobile 307
sources and machinery (13793 Mg y-1
), combustion in energy and transformation industries 308
(7322 Mg y-1
) and non-industrial combustion plants (7187 Mg y-1
), while remaining 309
EMEP/EEA sources (production processes, agriculture, waste treatment and disposal, solvent 310
and other product use, extraction and distribution of fossil fuels and geothermal energy and 311
14
other sources and sinks) accounted for 3216 Mg y-1
. The annual average levels of SO2 were 312
very low at all the sites and well below the European limit value. The EI reported that in 313
2007/8 the main contributors in Veneto were (in Mg y-1
): combustion in energy and 314
transformation industries (5077)> combustion in manufacturing industry (4578)> other 315
mobile sources and machinery (2340)> production processes (1879)> non-industrial 316
combustion plants (1327)> sum of other EMEP/EEA sources (165). 317
318
It should be noted that most of the NOx was emitted at ground level by mobile sources, 319
whereas most of SO2 emissions originated from stationary sources via chimneys. SO2 may 320
disperse widely from elevated sources, but the NOx sources are themselves widely 321
distributed. 322
323
3.2 Seasonal Variations 324
The PM2.5 time series are reported as Supplementary Information Figure SI1 and exhibit 325
seasonal trends at all of the sites, i.e. higher levels during winter and lower in summer, as 326
commonly observed in most sites in the Po Valley (e.g., Marcazzan et al., 2003; Vecchi et al., 327
2004; Perrone et al., 2012; Tositti et al., 2014). The seasonality is strongly linked to weather 328
conditions, such as prolonged atmospheric stability, shallower mixing layers, wind calm 329
periods and low temperatures, which favor the accumulation of atmospheric pollutants at the 330
ground level (Ferrero et al., 2010). The increased use of wood for domestic heating in winter 331
and the burning of biomass such as straw and crop residues in the harvest season (late 332
autumn) may also have a role in raising the PM2.5 levels. The semi-volatility of ammonium 333
nitrate may also be important. The time series also showed a number of single peaks of 334
concentration at various sites. In most cases these peaks occurred at individual sites and were 335
therefore linked to local and occasional phenomena. However, it is evident that the highest 336
15
concentrations were recorded on January 6th
for all stations except BL, when thousands of 337
folk fires of wooden material were lit in most of the Veneto region for a local religious 338
celebration. This episode was extensively reported by Masiol et al. (2014b) and recorded 339
extremely high daily concentrations of PM2.5, ranging from 136 µg m–3
in VI to 202 µg m–3
in 340
RO. This period was not included in the present study. 341
342
On a monthly basis, each ion exhibited a typical seasonality and similar seasonal trends were 343
generally observed in all the territory. Figure 2 reports the mass concentration time series of 344
the three SIA components, while seasonal average levels for all ions are shown as Figure SI2. 345
Results for SIA ions show that both the concentrations and the daily variations of SIA at the 346
four sites in the flatter areas of the Po Valley (VI, VE, PD and RO) are quite similar and are 347
in line with results observed at urban sites in other nearby regions (Table SI2). Low SIA 348
concentrations were recorded at all sites in June, and in October in the Alpine valley (BL), 349
the SIA components were extremely low as well. Nitrate concentrations in PM are inversely 350
related to the ambient temperature: they are higher in the colder months, mainly because 351
ammonium nitrate tends to volatilise at temperatures above 20°C (Schaap et al., 2004a; 352
Vecchi et al., 2009). This is observed all over Europe (e.g., Allen et al., 1989; Schaap et al., 353
2004b; Revuelta et al., 2012). Sulphate presents a peculiar bimodal seasonality, with two 354
maxima in August and February. A peak in the warmest period is commonly recorded in 355
Europe (e.g., Revuelta et al., 2012) and is probably due to the increased photochemical 356
activity favouring the oxidation of SO2 via hydroxyl radical reaction (Stockwell and Calvert, 357
1983; Khoder, 2002; Seinfeld and Pandis, 2006), whereas the peak in February may be 358
associated with aqueous phase oxidation. Ammonium concentrations tend to parallel those of 359
nitrate and sulphate. Calcium shows no evident seasonality. However slightly higher levels 360
were recorded in August and winter. Potassium, a known tracer of biomass combustions, 361
16
(e.g., Puxbaum et al., 2007; Saarnio et al., 2010) presents an evident seasonality with higher 362
concentrations in the coldest period. Wood (i.e. logs, briquettes, chips and pellet) is becoming 363
a popular renewable alternative to natural gas in Northern Italy (Pastorello et al., 2011) and 364
the increasing emissions from its use for domestic heating can be considered the most 365
plausible source. Chloride has a seasonal behavior similar to potassium. Its presence in PM 366
can derive from various sources, i.e. sea-salt, biomass burning, resuspension of road deicing 367
salts, coal combustion and various industrial processes. The marine origin can be probably 368
excluded as no significant gradients of concentration are observed from the stations close to 369
the coast (VE) to the more continental ones (VI and PD). Therefore, biomass burning and the 370
resuspension of road salt are probably the most important sources. 371
Seasonal trends of gaseous pollutants are also given in Figure SI2. Nitrogen oxides increased 372
during the cold season due to changes in mixing depths and emission rates, while ozone 373
reached the highest levels in the warmest period due to its photochemistry. Sulfur dioxide 374
showed no clear seasonal trends, but reached the highest levels in VE during the warmest 375
period (June-August). 376
377
3.3 Spatial Variations 378
Starting from the evidence that PM2.5 and most ions have quite similar seasonal trends at all 379
the sites, an inter-site comparison of the annual concentrations was conducted for each ion. 380
Since the data were not distributed normally, the nonparametric Kruskal–Wallis one-way 381
analysis of variance was used. This test is based on the rank of each sample instead of its 382
value and the null hypothesis assumes that the central values of the groups (medians) are 383
equal, and is rejected for p< 0.05. Thus, the post hoc Dunn’s test was applied to identify 384
which sites are significantly different from the others. Results generally show that PM2.5, 385
nitrate, sulphate and ammonium in BL and TV are significantly (p<0.05) different from the 386
17
other sites and concentrations increased from North (mountain) to South (lowland) and from 387
East (coastal) to West (more continental). On the other hand, K+ and Ca
2+ levels are not 388
significantly different and their concentrations are therefore uniform in all of the Veneto 389
region. These results show that biomass burning, which has been identified as a major source 390
of potassium, and the re-suspension of mineral dust and soil, which is the major source of 391
calcium, are quasi-uniformly distributed throughout the region. 392
393
An indirect quantification of differences in concentrations among the sites was carried out by 394
regressing PM2.5 mass concentration and nitrate+sulphate (expressed as neq m-3
) among 395
pairs of sites (intercept forced to zero). Results are provided in Figures SI3 and SI4, 396
respectively. Results for both PM2.5 and nitrate+sulphate show that sites located in the main 397
Po Valley (VI, VE, PD and RO) have regression slopes around 1 (0.84―1.17) and high 398
coefficients of determination (R2>0.8), which indicate good agreement between 399
concentrations. On the contrary, slopes (range 1.26―1.42) and R2 (≤0.2) between BL and 400
sites in the main Po Valley indicate a very poor agreement. TV has an intermediate behavior 401
with sites in the main Po Valley: it presents a moderate relationship (R2 0.6―0.8), but high 402
slopes (1.2―1.8). 403
404
The spatial and temporal relationships among the sites for PM2.5 and ionic species were 405
further investigated by using correlation analysis. A preliminary inter-site correlation analysis 406
among the PM2.5 concentrations for the whole year (365 days) was conducted. The PM2.5 407
distributions were tested for normality by applying the Shapiro-Wilk’s tests and the normality 408
assumption at p< 0.05 was not met. A Box-Cox transformation of the dataset was therefore 409
made. The resultant transformed data were normally distributed and Pearson’s correlation 410
analysis was run. Results (Table 3) generally show significant correlations (p<0.01, r>0.8) 411
18
among all the sites, with the exception of BL, which appears slightly less correlated (r≈0.7) 412
with the others. PM2.5 exhibits a similar temporal trend in all the cities even if these are 413
located in different territories of the region. It is evident that the processes of emission, 414
accumulation and removal are quite similar in the six cities. 415
416
However, the correlation analysis for the full dataset may be affected by the marked 417
seasonality of the variables, with the result that the correlation links variables with similar 418
seasonal trends and not sites with simultaneous daily variations. This problem was solved by 419
monthly-averaging the original data: the monthly means were subtracted from each daily 420
value in each selected period. This normalisation procedure had also the advantage of 421
generating variables that were quasi-normally distributed. The correlation matrices of the 422
monthly-averaged data are reported in Table 3 and show that the PM2.5 is still strongly 423
correlated at all the sites located in the lowland area, while the mountain site (BL) is less 424
correlated. Sulphate has usually significantly (p< 0.05) positive relationships for all pairs of 425
sites, indicating that it has a similar (synchronous) behavior in the whole region. Highly (r> 426
0.75) significant correlations are also found for PM2.5, nitrate and ammonium, except at BL 427
which appears to be uncorrelated with the other sites. Potassium is very well correlated in the 428
central part of the region (VI, VE, PD), while significant but weak correlations are found in 429
TV and RO, and BL is uncorrelated with the other sites. Calcium shows few inter-site 430
correlations (Table SI3). 431
432
This analysis generally shows that PM2.5, potassium, nitrate, sulphate and ammonium follow 433
a similar day-to-day trend at all sites throughout the region, in particular in the lowland 434
territory, and confirms that both the emission sources and the accumulation/removal 435
19
processes in the region are similar. A similar finding was also recently reported for the levels 436
of PM10-bound polycyclic aromatic hydrocarbons at 21 sites in Veneto (Masiol et al., 2013). 437
438
4. DISCUSSION 439
The SIA mass is generally calculated as the simple sum of ammonium, sulphate and nitrate or 440
is derived from the results of source apportionment approaches. Nevertheless, its prediction is 441
not straightforward because the ion generation, transport, aging or removal in the particle-442
phase strongly depends on weather conditions, but also on the presence of precursor gases 443
and oxidant species (mainly hydroxyl radical, hydrogen peroxide and ozone). Basically, SIA 444
generation is a two-step process, in which the gaseous precursors SO2 and NOx undergo 445
photochemical and heterogeneous thermal oxidation to form sulfuric and nitric acids, 446
respectively. Subsequently, the acids are neutralised by ammonia, and in the case of 447
ammonium nitrate, partitioned according to thermodynamic equilibria, mostly determined by 448
temperature and relative humidity (Baek et al., 2004; Seinfeld and Pandis, 2006; Allen et al., 449
1989). Reactions with other ions may also form mixed salts. Using the experimental data 450
obtained in this study, some preliminary conclusions regarding the SIA are drawn. 451
452
4.1 Sulfur and Nitrogen Oxidation Ratios 453
The degree of atmospheric conversion of gaseous precursors, SO2 and NO2, to sulphate and 454
nitrate, respectively, can be indirectly assessed by means of the sulfur (SOR) and nitrogen 455
(NOR) oxidation ratios: 456
SOR =nnssSO4
2−
nnssSO42− + nSO2
NOR =nNO3
−
nNO3− + nNO2
20
where the n units are in moles m–3
and nss-SO42 is the non-sea-salt sulphate calculated as 457
[SO42-
]- 0.25∙[Na+]. The SOR and NOR have been used by many authors (e.g., Khoder, 2002; 458
Bencs et al., 2008; Behera and Sharma, 2010) to describe the degree of ageing of the air 459
mass. The results appear as Table SI4, alongside those of other similar studies for 460
comparison. Annually, the average SOR varied from 0.4 (VE) to 0.6 (PD) suggesting a high 461
degree of oxidation of SO2 in the atmosphere, while the annual average NOR ranged between 462
0.04 (BL) and 0.1 (PD). SOR shows no clear spatial variation and generally its seasonal 463
concentrations follow those of sulphate. However, it is important to point out that the 464
minimum SOR is reached at VE in the warmest period. This is probably due to the highest 465
concentrations of SO2 in summer caused by: (1) the peak of energy production of a coal-fired 466
power plant meeting the demand for air conditioning; (2) the presence of higher shipping 467
traffic using the cruise harbour. This assumption is also supported by the emission inventory 468
for 2010 (ISPRA, 2014) showing that the Venice province has the highest production of SO2 469
(4586 Mg y-1
), followed by Padova (1324 Mg y-1
). About 71% of the emissions in VE are 470
attributed to combustion in energy and transformation industries. Spatially, NOR seems to 471
increase slightly from North to South and from the coast to the mainland. 472
473
4.2 Ammonia Availability and Neutralisation Ratio 474
Ammonia is known to neutralise sulfuric acid irreversibly, and then nitric acid. In addition, 475
hydrochloric acid may react with gaseous ammonia to form ammonium chloride aerosol. 476
However, in thermodynamic equilibrium conditions ammonium chloride is reported to be 2-3 477
times more volatile than ammonium nitrate (Stelson and Seinfeld, 1982; Pio et al., 1992) and 478
its formation occurs later. It is well known that in low ammonia conditions, NH3 acts as the 479
main limiting factor for SIA generation (Erisman and Schaap, 2004). On the other hand, in 480
case of high NH3 availability, ammonium nitrate formation is principally limited by the 481
21
availability of nitric acid. These conditions are important in agricultural areas because 482
livestock farming and the use of soil fertilizers are primary sources of atmospheric NH3 483
(Galloway et al., 2004; Sutton et al., 2008). Recent modeling simulations on a continental 484
scale (Wichink Kruit et al., 2012) have reported that ammonia levels in the Po Valley are 485
among the highest in Europe (range 4−10 μg m−3
). This is also confirmed by satellite 486
observations (Clarisse et al., 2009) indicating the Po valley as one of the most evident 487
hotspots for NH3 at a global scale. The 2010 Italian emission inventories (ISPRA, 2014) 488
reported that ~50.2·103 Mg of NH3 are emitted annually in Veneto, most of which is from 489
agriculture (48.9·103 Mg), followed by road transport (0.7·10
3 Mg). Because SO2 emissions 490
have been sharply reduced in the last decades in most developed countries, including Italy 491
(Manktelow et al., 2007; Hamed et al., 2010), more NH3 is available for the formation of 492
ammonium nitrate (Bauer et al., 2007; Pye et al., 2009). Recent data indicated that in Veneto 493
SO2 concentrations are generally < 8 μg m−3
, i.e. below the EU lower threshold (ARPAV, 494
2013). 495
496
Reactions of gaseous acids with other particles (e.g., sea salt, crustal dust, anthropogenic) can 497
form secondary salts, mainly replacing Cl– with sulphate and nitrate, or forming salts with 498
Na+, K
+, Mg
2+ or Ca
2+. For example, sulphate and nitrate may affect the hygroscopic 499
behaviour of mineral dust (Shi et al., 2008) and may form nitrate-containing particles mainly 500
in the coarse mode (Pakkanen et al., 1996; Metzger et al., 2006). 501
502
However, in this study, the masses of Na+, Mg
2+, Ca
2+ and Cl
– were low, if compared to 503
NH4+, NO3
– and SO4
2– and therefore their contribution to salts in PM2.5 can be assumed to be 504
negligible. From a linear regression analysis between ammonium and the sum of nitrate and 505
sulphate (expressed as neq m–3
) significant coefficients of determination (R2 varying from 506
22
0.94 in BL and 0.99 in VI, VE and RO), almost unitary slopes (from 0.83 in BL 1.06 in VI, 507
VE and RO) and very low intercepts were obtained for all the sites. The scatterplots are 508
reported as Figure SI5. They also reveal that the relationships are constantly linear in all the 509
seasons, even if the mass contributions of each ion varied greatly during the year. 510
511
The neutralisation ratio (NR) (Bencs et al., 2008), also called acidity ratio (Engelhart et al., 512
2011), expresses the degree of neutralisation of sulphate and nitrate by ammonium 513
(concentrations are in equivalents) and was used to describe the aerosol acidity: 514
NR =[NH4
+]
[SO42−] + [NO3
−]
Figure 3 shows the NR time series and permits some inferences: (i) on an annual basis, 515
average NRs were equal to 1 within the analytical variability, or slightly less: 0.8 in BL, TV, 516
PD and 0.9 in VI, VE, RO; (ii) the lowest NRs were recorded in spring, while they were 517
almost constant in the remaining months at all the sites; (iii) both the concentrations and the 518
daily variations of SIA at the 4 sites in the Po valley had similar trends; (iv) NR variability in 519
August and February, i.e. in the warmest and coldest months of the year, respectively, was 520
small, while strong daily changes were recorded in April and October. It is unclear if this 521
trend is linked mainly to a discontinuity of the sources (e.g., domestic heating switching off 522
and on), to weather factors controlling the SIA generation, or to external transport effects. 523
524
To investigate the extent of neutralisation of the SIA in more detail, NR was plotted against 525
the ammonium concentration (Figure 4a). Results show that for all the sites: when 526
concentrations of NH4+ exceed ~150 neq m
-3, the NR appears to be constant around 1 and 527
SIA is likely to be composed of ammonium nitrate and ammonium sulphates; for lower levels 528
of ammonia, the variability of NR increases and, generally, the ratio becomes smaller. These 529
23
results confirm that ammonia may effectively act as a limiting agent for SIA and suggest that 530
during ammonia-limiting conditions, sulfuric and nitric acids may react with other particles to 531
form salts. This assumption can be further confirmed by plotting the NR versus ionic balance 532
(ratio between the sum of all analysed cations and anions) (Figure 4b). The graph clearly 533
shows that most of samples are set in the 4th
quadrant, a region where the relative lack of 534
ammonium (NR<1) corresponds to an excess of cations (cations>anions), i.e. nitrate and 535
sulphate are potentially combined with other cations than ammonium. Figure 4b also shows 536
that no samples are plotted in the opposite quadrant (2nd
), demonstrating that on days with an 537
excess of ammonium (NR>1) no excess anions are present, thus showing the absence of other 538
inorganic salts of ammonium, such as NH4Cl. A few samples mainly pertaining to the 539
mountain site (BL) are scattered in the 1st and 3
rd quadrants: samples in the 1
st quadrant are 540
characterised by an excess of ammonium and a positive ionic balance, i.e. an excess of 541
positive charges probably neutralised by organic acids, not measured in this study. Samples 542
in the 3rd
quadrant were almost all collected in April and a possible explanation is that the 543
lack of positive charges may be balanced by H+ (which was not measured), resulting in acid 544
aerosol. 545
546
4.3 Potential Contribution of Long-Range Transport 547
The analysis of the back-trajectories was used to give some insight into the potential 548
contribution of long-range aerosol transport upon the Veneto region. As known from the 549
literature, the use of trajectories has some limitations in accuracy for various reasons (e.g., 550
Stohl et al., 1998). However, taking into account the range of associated uncertainties, the use 551
of some trajectory statistical methods is recognised as very useful to investigate potential 552
source areas (Kabashnikov et al., 2011; Abdalmogith and Harrison, 2005). 553
24
For the purpose of this study, the variability of back-trajectories was tested using different 554
starting heights and hours: errors associated with a single trajectory were reduced by 555
simulating four trajectories for each sampling day (at 6, 12, 18, 24 local time). The cluster 556
analysis was applied to all the 4-days back-trajectories computed and for the each site, i.e. 4 557
trajectories every day, which have been merged with daily data. In fact, this expedient 558
allowed the spread of daily chemical data over 4 trajectories and thus can account for days 559
that may have changes in trajectories within 24-h. The number of extracted clusters was 560
carefully evaluated by analysing the change in the total spatial variance and the best 561
compromise was 5 clusters for all the sites. Results show that all sites present similar mean 562
trajectories (Figure 5) named (1) Western Europe, (2) Mediterranean, (3) local, (4) Northern 563
Europe and (5) Eastern Europe. Statistics for chemical composition data in each cluster are 564
presented as boxplots in Figures 5 and SI6. The number of trajectories grouped in each 565
cluster generally differ among BL, TV and other sites (Table SI5). The reason is linked to the 566
topography of the territory: BL is located in an alpine valley, TV is at the border of Alps, 567
whereas other sites are located in flatter areas of the Po Valley. As a consequence, results for 568
BL and TV differ from the other sites with results sometimes showing opposite trends. 569
Generally, PM2.5, nitrate and K+ show similar results, with concentrations higher for cluster 3 570
in BL and TV and for clusters 1, 4 and 5 for the remaining sites. Sulphate in BL and TV 571
appears to have higher concentrations when air masses are associated with clusters 1 and 5, 572
whereas it is associated with clusters 3 and 5 at the other sites. Calcium and chloride show 573
only small differences. Sites BL and TV show a different behaviour with respect to PM2.5. 574
The highest concentrations are associated with trajectory 3, which for the other sites shows 575
the lowest PM2.5. This effect is probably the result of trapping of lower level emissions at BL 576
and TV in the slow moving northerly air. With regard to high sulphate and nitrates, these 577
sites behave rather similarly to the others as a results of regional influences. 578
25
579
On the other hand, the analysis of the potential effects of long-range transport on a regional 580
scale through the CWT model returned very similar results at all sites and clearly indicate 581
some predominant source areas for potential transboundary transport of PM2.5 and some ions 582
(Figures 6a and 6b). In particular, a wide area spanning across Eastern and Central Europe 583
and Northern Italy is identified as a main potential source of all species. Similar results have 584
also been obtained from a previous study conducted at a site near VE during 2009-2010 585
(Squizzato et al., 2014). In addition, other minor source areas are also identified: an area in 586
Central Italy which roughly coincides with the heavily populated areas of Rome and Naples 587
as a source of PM2.5 nitrate, and an area in North Africa, which may be linked to Saharan dust 588
outbreaks. CWT also shows that air masses passing over continental Europe are responsible 589
for the highest NR and SOR, while this effect is less evident for NOR (Figure SI7). If NR, 590
SOR and NOR are taken as indicative of the aging of air masses (generally highest values of 591
oxidation ratios and NR values close to 1 are expected in aged air masses) these results stress 592
that transboundary transport from continental Europe may have an important impact on levels 593
of secondary species in the Po Valley. The lower values of NOR than SOR probably reflect 594
the higher local emissions of NOx compared to SO2. 595
596
4.4 Analysis of Single Episodes 597
Three episodes of high SIA concentrations occurred during the campaign (Figure 2): (1) 15th 598
to 21st October, (2) 13th to 17th February and (3) 17th to 22nd February. Despite all sites 599
showed covariant daily variations in the levels of nitrate, some differences during those 600
episodes were identified. A further analysis of single back-trajectories was thus performed to 601
explain why some sites are generally experiencing much lower concentrations than others. 602
Figure 7 shows the single back-trajectories associated with the daily concentration of SIA. In 603
26
the first and third episodes, it is evident that all sites show similar daily air mass pathways 604
from the Mediterranean and Central Europe, respectively. However, only VI, VE, PD and RO 605
show similar daily variations and levels associated with single trajectories, whereas TV had a 606
similar daily variation, but significantly lower concentrations. There will be a number of 607
reasons explaining this result: (i) data indicate that transboundary transport of polluted air 608
masses may have a higher impact over the Eastern Po Valley; (ii) the cluster and CWT 609
analyses both indicate Central Europe as a major source area of ammonium nitrate aerosol; 610
(iii) results suggest the topography may influence the local impact of long-range transport: a 611
general homogeneity in the SIA levels is often recorded in the flat area of the valley, while 612
the Alpine chain may act as a barrier for the dispersion of pollutants at ground-level. 613
614
The results for the second episode are quite different. Despite all sites show similar air mass 615
histories, the levels of SIA were higher in RO and VI. As the differences cannot be explained 616
by differing air mass origins, it can be concluded that ammonium nitrate generation may also 617
occur locally as a consequence of oxidation of locally emitted NOx. 618
619
In conclusion, these results indicate that SIA pollution may be sensitive to both long-range 620
transport and local generation processes. Due to the relatively short period investigated in this 621
study (60 days over one year), there is a limit to the conclusions which may be drawn. 622
However, as a few events such as those considered in detail can have a considerable effect 623
upon the annual mean PM2.5 concentration, the different characteristics and effects of long-624
range or local SIA episodes should be investigated in more detail over a longer period, by 625
collecting a large number of samples. 626
627
628
27
5. CONCLUSIONS 629
This study is the first one investigating the spatial and temporal properties of secondary 630
inorganic aerosol in a large area of the Po Valley using simultaneous experimental 631
measurements at multiple receptor-sites. The statistical processing of the data shows that 632
PM2.5 and individual ions to have very similar concentrations across all urban sites and to be 633
very well correlated throughout the region, even though the sampling stations are located in 634
different cities and in an area ~18.4∙103 km
2-wide. Therefore, it can be concluded that the PM 635
pollution and the relative amount of SIA in the Veneto is quasi-uniformly distributed 636
throughout the region and the formation and removal processes affecting all sites are quite 637
similar. Moreover, a comparison with previous studies conducted in other nearby regions of 638
NE Italy indicates quite constant levels, seasonal trends and speciation of SIA over a wide 639
area of the Po Valley. The main results can be summarised as follows: 640
641
Annually, water soluble inorganic ions account from 30% to 41% of the total PM2.5 642
mass concentrations and the most abundant ion is nitrate (36%−47%), followed by 643
sulphate (22%−29%), ammonium (17%−21%) and potassium (3%−5%). 644
Each ion exhibits a characteristic seasonality and similar seasonal trends are generally 645
recorded over the entire study area. 646
PM2.5, nitrate, sulphate and ammonium in BL and TV are significantly different from 647
other sites and generally levels of analysed pollutants increased from North (mountain) 648
to South (lowland) and from East (coastal) to West (more continental). In contrast, K+ 649
and Ca2+
show weak spatial gradients. 650
Potassium, nitrate, sulphate and ammonium also show similar daily trends throughout 651
the region, in particular in the lowland territory, and confirm that both the sources and 652
the accumulation/removal processes in the region are similar. 653
28
The neutralisation ratio and the ionic balance were jointly investigated to provide 654
information about the processes affecting SIA and the interactions between the 655
secondary ions and other particles. Results confirm the probable formation of 656
secondary salts with potassium, sodium and calcium. 657
The application of trajectory-based methods (cluster and CWT analyses) was useful to 658
identify potential source areas leading to increases in PM2.5 and ions concentrations 659
across the region. Results showed that higher concentrations of all analysed species are 660
mainly associated with air masses originating in a widespread area located in the 661
Eastern-Central Europe. Central Italy and Northern Africa are also identified as 662
possible source areas particularly for PM2.5 and K+. 663
The analysis of three single episodes of high ammonium nitrate levels indicate that both 664
long-range transport and local formation processes may lead to high SIA levels during 665
colder months. Those events have a large potential for raising the annual average levels 666
of PM2.5 and should be investigated in more detail. 667
668
As a final remark, this study concluded that SIA pollution has similar and concurrent effects 669
over the entire study area and probably in the whole Po Valley. Findings clearly indicate that 670
any action to mitigate the PM2.5 pollution to meet the present target and the future air quality 671
standards in Veneto must be taken concurrently in the entire region and well beyond its 672
boundaries. 673
674
ACKNOWLEDGEMENTS 675
This study was conducted within an agreement between the Ca’ Foscari University of Venice 676
and ARPAV. The authors gratefully acknowledge the NOAA Air Resources Laboratory 677
29
(ARL) for the provision of the HYSPLIT transport and dispersion model and/or READY 678
website (http://www.ready.noaa.gov) used in this publication. 679
680
DISCLAIMER 681
This study was not financially supported by any public or private institution. We would like 682
to stress that the views expressed in this study are exclusively of the authors and do not 683
necessarily correspond to those of ARPAV. 684
685 686
687
30
REFERENCES 688
Abdalmogith S.S., Harrison R.M., 2005. The use of trajectory cluster analysis to examine the 689
long-range transport of secondary inorganic aerosol in the UK. Atmospheric Environment 39, 690 6686-6695. 691 692 Allen, A. G., Harrison, R.M., Erisman, J. W., 1989. Field measurements of the dissociation 693 of ammonium nitrate and ammonium chloride aerosols. Atmospheric Environment 23, 694
1591-1599. 695 696 Andreani-Aksoyoglu, S., Prévôt, A. S. H., Baltensperger, U., Keller, J., Dommen, J., 2004. 697 Modeling of formation and distribution of secondary aerosols in the Milan area (Italy). 698
Journal of Geophysical Research 109 (D5), D05306. doi:10.1029/2003JD004231. 699 700 Appel, B.R., Tokiwa, Y., Haik, M., Kothny, E.L., 1984. Artifact of particulate sulfate and 701 nitrate formation on filter media, Atmospheric Environment 18, 409–416. 702
703 ARPAV (Environmental Protection Agency of Veneto Region), 2013. Regional Report of Air 704 Quality–Year 2012, pp. 85 [in Italian]. Available at: http://www.arpa.veneto.it/temi-705
ambientali/aria/riferimenti/documenti 706 707 Baek B. H., Aneja V. P., Tong Q., 2004. Chemical coupling between ammonia, acid gases, 708
and fine particles. Environmental Pollution 129, 89–98. 709 710
Bauer, S. E., Koch, D., Unger, N., Metzger, S. M., Shindell, D. T., Streets, D. G., 2007. 711 Nitrate aerosols today and in 2030: a global simulation including aerosols and tropospheric 712
ozone. Atmospheric Chemistry and Physics 7, 5043-5059. 713 714
Behera S. N., Sharma M., 2010. Investigating the potential role of ammonia in ion chemistry 715 of fine particulate matter formation for an urban environment. Science of the Total 716 Environment 408, 3569-3575 717
718
Bencs, L., Ravindra, K., de Hoog, J., Rasoazanany, E. O., Deutsch, F., Bleux, N., Berghmans, 719 P., Roekens, E., Krata, A., Van Grieken, R., 2008. Mass and ionic composition of 720 atmospheric fine particles over Belgium and their relation with gaseous air pollutant. Journal 721 of Environmental Monitoring 10, 1148–1157. 722
723 Benson D. R., Yu J. H., Markovich A., Lee S.-H., 2011. Ternary homogeneous nucleation of 724 H2SO4, NH3, and H2O under conditions relevant to the lower troposphere. Atmospheric 725
Chemistry and Physics 11, 4755-4766. 726 727 Carslaw D. C., Ropkins K., 2012. openair - an R package for air quality data analysis. 728 Environmental Modelling & Software 27-28, 52-61. 729 730
Carslaw D. C., 2014. The openair manual — open-source tools for analysing air pollution 731 data. Version 10th June 2014, King’s College London. 732 733 Cheng, Y.H. Tsai, C.J., 1997. Evaporation loss of ammonium nitrate particles during filter 734
sampling. Journal of Aerosol Science 28, 1553–1567. 735 736
31
Clarisse, L., Clerbaux, C., Dentener, F., Hurtmans, D., Coheur, P. F., 2009. Global ammonia 737 distribution derived from infrared satellite observations. Nature Geoscience 2, 479-483. 738 739 Dasch, J.M., Cadle, S.H., Kennedy, K.G., Mulawa, P.A., 1989. Comparison of annular 740 denuders and filter packs for atmospheric sampling. Atmospheric Environment 23, 2775–741
2782. 742 743 Decesari, S., Allan, J., Plass-Duelmer, C., Williams, B. J., Paglione, M., Facchini, M. C., 744 O’Dowd, C., Harrison, R. M., Gietl, J. K., Coe, H., Giulianelli, L., Gobbi, G. P., Lanconelli, 745 C., Carbone, C., Worsnop, D., Lambe A. T., Ahern, A. T., Moretti, F., Tagliavini, E., Elste, 746
T., Gilge, S., Zhang, Y., Dall’Osto, M., 2014. Measurements of the aerosol chemical 747 composition and mixing state in the Po Valley using multiple spectroscopic techniques. 748
Atmospheric Chemistry and Physics 14, 12109-12132. 749 750 de Leeuw F.A.A.M., 2002. A set of emission indicators for long-range transboundary air 751 pollution. Environmental Science and Policy 5, 135-145. 752 753
Draxler R.R., Rolph G.D., 2013. HYSPLIT (HYbrid Single-Particle Lagrangian Integrated 754 Trajectory) Model access via NOAA ARL READY Website 755
(http://www.arl.noaa.gov/HYSPLIT.php). NOAA Air Resources Laboratory, College Park, 756 MD. 757
EEA (European Environment Agency), 2013. AirBasedThe European Air Quality Database. 758 Available from: http://www.eea.europa.eu/themes/air/airbase (last accessed January, 2013). 759 760
Engelhart, G. J., Hildebrandt, L., Kostenidou, E., Mihalopoulos, N., Donahue, N. M., Pandis, 761
S. N., 2011. Water content of aged aerosol. Atmospheric Chemistry and Physics 11, 911-920. 762 763 Erisman, J.W., Schaap M., 2004. The need for ammonia abatement with respect to secondary 764
PM reductions in Europe. Environmental Pollution, 129, 159-163. 765 766
Ferrero L., Perrone M.G., Petraccone S., Sangiorgi G., Ferrini B. S., Lo Porto C., et al., 2010. 767 Vertically-resolved particle size distribution within and above the mixing layer over the 768 Milan metropolitan area. Atmospheric Chemistry and Physics 10, 3915-3932. 769
770 Galloway, J. N., Dentener, F. J., Capone, D. G., Boyer, E. W., Howarth, R. W., Seitzinger, S. 771
P., Asner, G. P., Cleveland, C. C., Green, P. A., Holland, E. A., Karl, D. M., Michaels, A. F., 772 Porter, J. H., Townsend, A. R., Vöosmarty, C. J., 2004. Nitrogen Cycles: Past, Present, and 773
Future. Biogeochemistry 70, 153-226. 774 775 Hamed, A., Birmili, W., Joutsensaari, J., Mikkonen, S., Asmi, A., Wehner, B., G. Spindler, 776 Jaatinen, A., Wiedensohler ,A., Korhonen, H., Lehtinen, K. E. J., Laaksonen, A., 2010. 777 Changes in the production rate of secondary aerosol particles in Central Europe in view of 778
decreasing SO2 emissions between 1996 and 2006. Atmospheric Chemistry and Physics 10, 779 1071-1091. 780 781 Harrison, R.M., Sturges, W.T., Kitto, A.M.N., Li, Y., 1990. Kinetics of evaporation of 782 ammonium chloride and ammonium nitrate aerosols. Atmospheric Environment 24A, 1883–783
1888. 784
785
32
Harrison, R.M., Kitto, A.M.N., 1990. Field intercomparison of filter pack and denuder 786 sampling methods for reactive gaseous and particulate pollutants. Atmospheric Environment 787 24A, 2633–2640. 788 789 Holmes N. S., 2007. A review of particle formation events and growth in the atmosphere in 790
the various environments and discussion of mechanistic implications. Atmospheric 791 Environment 41, 2183–2201. 792 793 Hsu, Y. K., Holsen, T. M., Hopke, P. K., 2003. Comparison of hybrid receptor models to 794 locate PCB sources in Chicago. Atmospheric Environment 37, 545-562. 795
796 ISPRA, 2014. Disaggregated national emission inventory 2010, Italian Institute for 797
Environmental Protection and Research, available online: 798 http://www.sinanet.isprambiente.it/it/sia-ispra/inventaria/versione-2.0-dell2019inventario-799 provinciale-delle-emissioni-in-atmosfera/view, last access: 1 September 2013. 800 801 Kabashnikov, V. P., Chaikovsky, A. P., Kucsera, T. L., Metelskaya, N. S., 2011. Estimated 802
accuracy of three common trajectory statistical methods. Atmospheric Environment, 45, 803 5425-5430. 804
805 Khoder M. I., 2002. Atmospheric conversion of sulfur dioxide to particulate sulfate and 806
nitrogen dioxide to particulate nitrate and gaseous nitric acid in an urban area. Chemosphere 807 49, 675-684. 808 809
Koutrakis, P., Thompson, K. M., Wolfson, J. M., Spengler, J. D., Keeler, G. J., Slater, J. L., 810
1992. Determination of aerosol strong acidity losses due to interactions of collected particles: 811 Results from laboratory and field studies. Atmospheric Environment 26A, 987–995. 812 813
Manktelow, P. T., Mann, G. W., Carslaw, K. S., Spracklen, D.V., Chipperfield, M. P., 2007. 814 Regional and global trends in sulfate aerosol since the 1980s. Geophysical Research Letter 815
34, L14803. 816 817 Marcazzan G. M., Ceriani M., Valli G., Vecchi R., 2003. Source apportionment of PM10 and 818
PM2.5 in Milan (Italy) using receptor modeling. Science of The Total Environment 317, 137-819 147. 820
821 Masiol M., Formenton G., Pasqualetto A., Pavoni B., 2013. Seasonal trends and spatial 822
variations of PM10-bounded polycyclic aromatic hydrocarbons in Veneto region, Northeast 823 Italy. Atmospheric Environment 79, 811-821. 824 825 Masiol M., Squizzato S., Rampazzo G., Pavoni B., 2014a. Source apportionment of PM2.5 at 826 multiple sites in Venice (Italy): Spatial variability and the role of weather. Atmospheric 827
Environment 98, 78-88. 828 829 Masiol M., Formenton G., Giraldo G., Pasqualetto A., Tieppo P., Pavoni B., 2014b. The dark 830 side of the tradition: the polluting effect of Epiphany folk fires in the eastern Po Valley 831 (Italy). Science of the Total Environment 473-474, 549-564. 832
833
33
Metzger, S., Mihalopoulos, N., Lelieveld, J., 2006. Importance of mineral cations and 834 organics in gas-aerosol partitioning of reactive nitrogen compounds: case study based on 835 MINOS results. Atmospheric Chemistry and Physics 6, 2549-2567. 836 837 Pakkanen, T. A., Kerminen, V. M., Hillamo, R. E., Makinen, M., Makela, T., Virkkula, A., 838
1996. Distribution of nitrate over seasalt and soil derived particles – Implications from a field 839 study. Journal of Atmospheric Chemistry 24, 189-205. 840 841 Pastorello C., Caserini S., Galante S., Dilara P., Galletti F., 2011. Importance of activity data 842 for improving the residential wood combustion emission inventory at regional level. 843
Atmospheric Environment 45, 2869-2876. 844 845
Pathak, R.K., Yao, X.H., Chan, C.K., 2004. Sampling artifacts of acidity and ionic species in 846 PM2.5. Environmental Science and Technology 38, 254–259. 847 848 Pathak, R.K. Chan, C.K., 2005. Inter-particle and gas-particle interactions in sampling 849 artifacts of PM2.5 in filter-based samplers. Atmospheric Environment 39, 1597–1607. 850
851 Perrone M. G., Larsen B. R., Ferrero L., Sangiorgi G., De Gennaro G., Udisti R., Zangrando, 852
R., Gambaro, A., Bolzacchini, E., 2012. Sources of high PM2.5 concentrations in Milan, 853 Northern Italy: Molecular marker data and CMB modeling. Science of The Total 854
Environment 414, 343-355. 855 856 Pio, C. A., Nunes, T. V., Leal, R. M., 1992. Kinetic and thermodynamic behaviour of volatile 857
ammonium compounds in industrial and marine atmospheres. Atmospheric Environment. 858
Part A. General Topics 26, 505-512. 859 860 Putaud, J.P., Van Dingenen, R., Raes, F., 2002. Submicron aerosol mass balance at urban and 861
semirural sites in the Milan area (Italy). Journal of Geophysical Research 107 (D22), 8198–862 8208. 863
864 Putaud J.-P., van Dingenen R., Alastuey A., Bauer H., Birmili W., Cyrys J., Flentje, H., 865 Fuzzi, S.,Gehrig, R., Hansson, H.C., Harrison, R.M., Herrmann, H., Hitzenberger, R., Hüglin, 866
C., Jones, A.M., Kasper-Giebl, A., Kiss, G., Kousa, A., Kuhlbusch, T.A.J., Löschau, G., 867 Maenhaut, W., Molnar, A., Moreno, T., Pekkanen, J., Perrino, C., Pitz, M., Puxbaum ,H., 868
Querol, X., Rodriguez, S., Salma, I., Schwarz, J., Smolik, J., Schneider, J., Spindler, G., ten 869 Brink, H., Tursic, J., Viana, M., Wiedensohler, A., Raes, F., 2010. A European aerosol 870
phenomenology – 3: Physical and chemical characteristics of particulate matter from 60 rural, 871 urban, and kerbside sites across Europe. Atmospheric Environment 44, 1308-1320. 872 873 Puxbaum H., Caseiro A., Sánchez-Ochoa A., Kasper-Giebl A., Claeys M., Gelencser A., 874 Legrand, M., Preunkert, S., Pio, C., 2007. Levoglucosan levels at background sites in Europe 875
for assessing the impact of biomass combustion on the European aerosol background. Journal 876 of Geophysical Research 112, D23S05. http://dx.doi.org/10.1029/2006JD008114. 877 878 Pye, H. O. T., Liao, H., Wu, S., Mickley, L. J., Jacob, D. J., Henze, D. K., Seinfeld, J. H., 879 2009. Effect of changes in climate and emissions on future sulfate-nitrate-ammonium aerosol 880
levels in the United States. Journal of Geophysical Research-Atmospheres 114, D01205, 881
doi:10.1029/2008jd010701. 882
34
883 Revuelta, M. A., Harrison, R. M., Núñez, L., Gomez-Moreno, F. J., Pujadas, M., Artíñano, 884 B., 2012. Comparison of temporal features of sulphate and nitrate at urban and rural sites in 885 Spain and the UK. Atmospheric Environment 60, 383-391. 886 887
Rolph, G.D., 2013. Real-time Environmental Applications and Display sYstem (READY) 888 Website (http://www.ready.noaa.gov). NOAA Air Resources Laboratory, College Park, MD. 889 890 Saarnio K., Aurela M., Timonen H., Saarikoski S., Teinila K., Makela T., Sofiev, M., 891 Koskinen, J., Aalto, P. P., Kulmala, M., Kukkonen, J., Hillamo, R., 2010. Chemical 892
composition of fine particles in fresh smoke plumes from boreal wild-land fires in Europe. 893 Science of the Total Environment 408, 2527-2542. 894
895 Schaap, M., Spindler, G., Schulz, M., Acker, K., Maenhaut, W., Berner, A., et al., 2004a. 896 Artefacts in the sampling of nitrate studied in the “INTERCOMP” campaigns of 897 EUROTRAC-AEROSOL. Atmospheric Environment 38, 6487-6496. 898 899
Schaap, M., van Loon, M., ten Brink, H. M., Dentener, F. J., Builtjes, P. J. H., 2004b. 900 Secondary inorganic aerosol simulations for Europe with special attention to nitrate. 901
Atmospheric Chemistry and Physics 4, 857-874. 902 903
Seibert, P., Kromp-Kolb, H., Baltensperger, U., Jost, D. T., Schwikowski, M., 1994. 904 Trajectory analysis of high-alpine air pollution data. In Air pollution modeling and its 905 application X. Springer US, pp 595-596. 906
907
Seinfeld, J. H., Pandis, S. N., 2006. Atmospheric Chemistry and Physics, second ed. In: From 908 Air Pollution to Climate Change John Wiley & Sons, NewYork. 909 910
Shi Z., Zhang D., Hayashi M., Ogata H., Ji H., Fujiie W., 2008. Influences of sulfate and 911 nitrate on the hygroscopic behaviour of coarse dust particles. Atmospheric Environment 912
42(4), 822-827. 913 914 Squizzato S., Masiol M., Brunelli A., Pistollato S., Tarabotti E., Rampazzo G., Pavoni B., 915
2013. Factors determining the formation of secondary inorganic aerosol: a case study in the 916 Po Valley (Italy). Atmospheric Chemistry and Physics 13, 1927-1939. 917
918 Squizzato, S., Masiol, M., Visin, F., Canal, A., Rampazzo, G., Pavoni, B., 2014. The PM2.5 919
chemical composition in an industrial zone included in a large urban settlement: main sources 920 and local background. Environmental Science: Processes & Impacts 16, 1913-1922. 921 922 Stelson A. W. Seinfeld D. H., 1982. Relative humidity and temperature dependence of the 923 ammonium nitrate dissociation constant. Atmospheric Environment 16, 983-992. 924
Stockwell, W. R., Calvert, J. G., 1983. The mechanism of the HO-SO2 reaction. Atmospheric 925 Environment 17, 2231-2235. 926 927 Stohl, A., 1998. Computation, accuracy and applications of trajectories-a review and 928 bibliography. Atmospheric Environment 32, 947-966. 929
930
35
Sutton, M. A., Erisman, J. W., Dentener, F., Möller, D., 2008. Ammonia in the environment: 931 From ancient times to the present. Environmental Pollution 156, 583-604. 932 933 Tositti L., Brattich E., Masiol M., Baldacci D., Ceccato D., Parmeggiani S., Stracquadanio, 934 M., Zappoli, S., 2014. Source apportionment of particulate matter in a large city of 935
southeastern Po Valley (Bologna, Italy). Environmental Science and Pollution Research 21, 936 872-890. 937 938 Vecchi, R., Marcazzan, G., Valli, G., Ceriani, M., Antoniazzi, C., 2004. The role of 939 atmospheric dispersion in the seasonal variation of PM1 and PM2.5 concentration and 940
composition in the urban area of Milan (Italy). Atmospheric Environment 38, 4437-4446. 941 942
Vecchi, R., Valli, G., Fermo, P., D’Alessandro, A., Piazzalunga, A., Bernardoni, V., 2009. 943 Organic and inorganic sampling artefacts assessment. Atmospheric Environment 43, 1713-944 1720. 945 946 WHO, 2006. Air Quality Guidelines, Global Update 2005. World Health Organisation, 947
Geneva. 948 949
Wichink Kruit, R. J., Schaap, M., Sauter, F. J., van Zanten, M. C., van Pul, W. A. J., 2012. 950 Modeling the distribution of ammonia across Europe including bi-directional surface–951
atmosphere exchange. Biogeosciences 9, 5261-5277. 952 953
Wu, S.-Y., Hu, J.-L., Zhang, Y., Aneja, V.P., 2008. Modeling atmospheric transport and fate 954
of ammonia in North Carolina e part II: effect of ammonia emissions on fine particulate 955 matter formation. Atmospheric Environment 42, 3437-3451. 956
957 Zhang, X. Q. McMurry, P.H., 1992. Evaporative loss of fine particulate nitrates during 958
sampling. Atmospheric Environment 26A, 3305–3312. 959
36
TABLE LEGENDS 960 961 962 Table 1. Characteristics of the selected sampling sites and the number of analysed 963
samples. 964 965
Table 2. Annual average concentrations of analysed pollutants. A full list of results 966 including monthly average concentrations is provided as supplementary 967 material Table SI1. 968
969
Table 3. Inter-site correlation matrices. Upper-left: box-cox transformed PM2.5 dataset 970 for the whole year (365 day); other matrices are calculated on the selected 971
periods (60 days) and data were monthly normalized. Only significant (p< 972
0.05) correlations are shown; correlations significant (p<0.01) are bold faced. 973 Correlation matrices for all analysed compounds is provided in Table SI2. 974
975 976 977
978 979
FIGURE LEGENDS 980 981
982 Figure 1. Map of selected sites (a; left) and annual average percentages of analysed ions 983
on ΣWSII (b; right). 984
985
Figure 2. Time series of sulphate, nitrate and ammonium in the six sites. 986
987
Figure 3. Time series of neutralisation ratio (NR) in the six sites. 988
989
Figure 4. Scatterplots of a) ammonium vs NR and b) ionic balance vs NR. Samples 990 collected in six sites are coloured differently. 991
992
Figure 5. Results of the back-trajectory clustering (upper) and distributions of PM2.5 and 993 ion concentrations for each identified cluster (bottom). Results for remaining 994 ions are provided as Supplementary Information Figure SI6. 995
996 Figure 6a. CWT analysis for PM2.5, nitrate and sulphate. Concentrations are expressed as 997
μg m‒3
. 998
999
Figure 6b. CWT analysis for chloride, potassium and calcium. Concentrations are 1000
expressed as μg m‒3
. 1001
1002
Figure 7. Single back-trajectories during three high-nitrate concentration events. 1003
1004
1005
37
Table 1. Characteristics of the selected sampling sites and the number of analysed samples. 1006
1007 1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
Municipality Latitude Longitude Alt (m) Site characteristics Other automatic measurements
BL Belluno 46.143 N 12.218 E 401 Public park, residential-commercial area SO2; O3; NO2 ; NO ; NOx ; CO ; Benzene ; PM10 (gravimetric); PM10 (BAMs)
TV Conegliano 45.890 N 12.307 E 72 Residential area SO2 ; O3 ; NO2 ; NO ; NOx ; CO ; PM10 (gravimetric)
VI Vicenza 45.560 N 11.539 E 36 Residential area NO2 ; NO ; NOx ; PM10 (gravimetric); PAHs
PD Padova 45.371 N 11.841 E 13 Residential area SO2 ; O3 ; NO2 ; NO ; NOx ; CO ; Benzene ; PM10 (gravimetric); PAHs
VE Venice-Mestre 45.498 N 12.261 E 1 Public park, residential area SO2 ; O3 ; NO2 ; NO ; NOx ; CO ; Benzene ; PM10 (gravimetric); PAHs
RO Rovigo 45.074 N 11.782 E 7 Residential-commercial area SO2 ; TSP (gravimetric); O3 ; NO2 ; NO ; NOx ; CO ; PM10 (gravimetric)
38
Table 2. Annual average concentrations of analysed pollutants. A full list of results including 1019 monthly average concentrations is provided as supplementary material Table SI1. 1020 1021
BL TV VI VE PD RO
PM2.5 µg m–3 17 20 28 25 29 27
Na+ µg m–3 0.14 0.31 0.15 0.16 0.47 0.23
NH4+ µg m–3 0.9 1.1 2.3 1.9 2 2.3
K+ µg m–3 0.28 0.29 0.31 0.38 0.39 0.3
Ca2+ µg m–3 0.11 0.15 0.15 0.15 0.16 0.15
Cl– µg m–3 0.12 0.12 0.19 0.17 0.19 0.24
NO3– µg m–3 1.8 2.4 5 3.6 4.6 5.2
SO42– µg m–3 1.5 1.7 2.4 2.6 2.4 2.6
SIA µg m–3 4.2 5.2 9.7 8.1 9 10.2
SIA % 23 25 32 30 29 35
ΣWSII µg m–3 5.2 6.5 10.8 9.2 10.7 11.4
ΣWSII % 30 34 36 35 38 41
NO µg m–3 15 12 24 22 27 26
NO2 µg m–3 23 27 33 32 37 36
NOx µg m–3 45 45 70 65 79 76
O3 µg m–3 49 47 48 49 61 46
SO2 µg m–3 1.1 — — 2.8 1 2.5
1022
39
Table 3. Inter-site correlation matrices. Upper-left: box-cox transformed PM2.5 dataset for the whole year (365 day); other matrices are
calculated on the selected periods (60 days) and data were monthly normalized. Only significant (p< 0.05) correlations are shown; correlations
significant (p<0.01) are bold faced. Correlation matrices for all analysed compounds is provided in Table SI2.
PM2.5
(Whole year)
BL TV VI VE PD RO
PM2.5
(Monthly norm.)
BL TV VI VE PD RO
BL 1
BL 1
TV 0.74 1
TV 0.33 1
VI 0.75 0.86 1
VI 0.26 0.84 1
VE 0.75 0.82 0.86 1
VE
0.84 0.89 1
PD 0.74 0.83 0.89 0.94 1
PD 0.29 0.85 0.89 0.87 1
RO 0.71 0.82 0.88 0.9 0.93 1
RO 0.26 0.8 0.81 0.83 0.95 1
NO3–
(Monthly norm.)
BL TV VI VE PD RO
SO42–
(Monthly norm.)
BL TV VI VE PD RO
BL 1
BL 1
TV
1
TV 0.51 1
VI
0.84 1
VI 0.39 0.51 1
VE
0.85 0.95 1
VE 0.53 0.86 0.58 1
PD
0.87 0.97 0.96 1
PD 0.54 0.74 0.59 0.9 1
RO
0.79 0.86 0.84 0.92 1
RO 0.39 0.73 0.55 0.83 0.89 1
NH4+
(Monthly norm.)
BL TV VI VE PD RO
K+
(Monthly norm.)
BL TV VI VE PD RO
BL 1
BL 1
TV
1
TV
1
VI 0.26 0.81 1
VI
0.58 1
VE
0.86 0.92 1
VE
0.64 0.82 1
PD
0.87 0.94 0.95 1
PD
0.51 0.83 0.77 1
RO 0.26 0.77 0.85 0.83 0.92 1 RO 0.32 0.52 0.56 0.77 1
40
Figure 1. Map of selected sites (a; left) and annual average percentages of analysed ions on ΣWSII (b; right).
41
Figure 2. Time series of sulphate, nitrate and ammonium in the six sites.
42
Figure 3. Time series of neutralisation ratio (NR) in the six sites.
43
Figure 4. Scatterplots of a) ammonium vs NR and b) ionic balance vs NR. Samples
collected in six sites are coloured differently.
44
Figure 5. Results of the back-trajectory clustering (upper) and distributions of PM2.5 and ion concentrations for each identified cluster
(bottom). Results for remaining ions are provided as Supplementary Information Figure SI6.
45
Figure 6a. CWT analysis for PM2.5, nitrate and sulphate. Concentrations are expressed as μg m
‒3.
46
Figure 6b. CWT analysis for chloride, potassium and calcium. Concentrations are expressed as μg m
‒3.
47
Figure 7. Single back-trajectories during three high-nitrate concentration events.
48
Graphical abstract